General > Off Topic

Finaly 3.0 is somewhat out (and stuff)

<< < (4/13) > >>

PurpleYouko:

--- Quote ---The problem (as I already explained it) is that the robot does not mutate new stuff at all
--- End quote ---

I don't get this!  :blink:

I never did get it!

In every test that I have ever run (and that would be thousands), robots can and do mutate new stuff all the time.

I even built the scripting system to prove it and that is just what it did.

Try putting a simple robot in the sim thae set up a bunch of scripts that pause the game when a robot mutates a new value (some command that isn't in the original). It may take a while but sooner or later one of the robots WILL evolve the new item.

I have performed this test loads of times and it always works, thus proving that robots do evolve new stuff.

Unless we are using different definitions of stuff

 :blink:  PY  :blink:

Numsgil:
Okay, here are some things again:

1.  DB will probably take longer than other AL sims to reach useful mutations because it is so open ended.  Darwinpond (another AL sims) only has like half a dozen different memory locations that can be mutated, each controlling something rather macroscopic about the bot.  In darwinbots, the individual commands are much more specific.

Also, until I get to creating a more realistic probablility distribution for one DNA command to turn into another, you are going to see [span style=\'font-size:30pt;line-height:100%\'][you]MOST[/you][/span] of the mutations be either bad or neutral.  Turning a refeye into a poison will generally break the genes.

2.  Okay, in the one quote where I said unmated, I meant unmutated.

3.  I've seen bots evolve that are better from the ancestors.  Heck, Carlo's original C_Ancestralis evolved into several different bots that were better than the original.  [span style=\'font-size:30pt;line-height:100%\']MUTATIONS DO WORK[/span].  If you can't get them to work than you are doing something [span style=\'font-size:30pt;line-height:100%\']WRONG[/span]

4. If you are not willing to actually wait for mutations and evolution to take hold I strongly recommend finding another AL simulator.  Don't try to say mutations don't work if you only give it half an hour.  The program does not say "gee, what mutation will be the most beneficial?"  Heck, the DNA of most successful bots is hundreds of elements long.  Do you have any idea how long that'd take to accumulate by natural mutations?

Please, before anyone says that mutations don't work, [span style=\'font-size:30pt;line-height:100%\']try running the program for most of an afternoon or even a few days[/span].

How many cycles do you run your mutation sims for?  2000?  Try 3 million cycles.  Then you'll see what I'm talking about.

Numsgil:
Sorry, teaches me to post right after I wake up (I'm not a morning person).

Here's a bot I mutated:


--- Code: ---'#name: Hallowers One Gene Bot.txt
'#generation: 28
'#mutations: 20

cond
  *.slime
  26289
  <
  *.nrg
  31996
  <
start
  10
stop

cond
start
  *.eye5
  10
  add
  mult
  7
  div
  sub
  7
  div
  sub
  .up
  store
  *.eye5
  *.eye1
  sub
  *.eye5
  *.eye2
  sub
stop

cond
start
  *.trefdx
  *.eye3
stop

cond
start
  inc
  sub
  *.eye5
  *.eye4
stop

cond
start
  sub
  add
  *.eye4
  sub
  rnd
  add
  add
  add
  4
  div
  *.eye5
  *.eye6
  sub
  *.mass
  rnd
  *.eye7
  sub
  *.eye5
  *.eye8
  sub
  *.eye5
  *.eye9
  sub
  add
  add
  add
  4
  div
  sub
  *.refeye
  1
  add
  mult
  5
stop

cond
start
  store
stop

cond
start
 -6
  7
  *.refup
  sub
  store
  50
  299
  *.ypos
  add
  store
stop
end

'#hash: q?_v/77lB|_r^;CvcNL6
--- End code ---

Lotsa stuff in there that's new.  Most of it doesn't make sense at all, but you can see that, yes, new stuff sometimes evolves.

Botsareus:
Nice lest post Num , ok ok , I am working on it...

Found it , (new post)

Botsareus:
There is a bug in the program that sayes that special mutation rates will do worse then normal once. If I give a bot in the begining its own mutation rates it developed in the lest run, It will do worse then the nonmutating bot. If I give defults witch are ("mofm:2 , 200 , 200 , all 200 , etc.") it will do better then the nonmutating bot.

This mean eather there is a bug or there is some rule of nature that says that defult rates are required for every new match.

At any rate, I have basicaly proven the theory of evolution and now working evolving a bot that can beat Din and the rest of them.

The only problem now is that ("History Repeats Itself") a bot that evolved now does worse agenst a bot 4 runs back then agenst a bot that it evolved agenst. The solution is to save each step of evolution and then evolve the robot agenst multipule bots. That makes me wonder: How well will humens do if the Dinasors are back?

Navigation

[0] Message Index

[#] Next page

[*] Previous page

Go to full version